Critical Thinking Partner
Four pieces from one session. The session started with a problem — AI is trained to agree with you — and ended with two finished artifacts: a Critical Partner skill with a recursive audit trail, and a Grounded Challenger voice profile born from a throwaway rewrite request.
The pieces don’t need to be read in sequence. Read what matches the problem you’re working on.
Contents
| Article | What it covers |
|---|---|
| The Most Dangerous Thing AI Does Is Agree With You | The sycophancy mechanism, why “play devil’s advocate” fails, and seven specific override instructions that actually change the model’s behavior |
| Your AI Is Lying to You. Politely. Every Single Day. | The same argument in a sharper voice — the Grounded Challenger in action, written for the reader who needs the punch before the explanation |
| I Made the Skill Audit Itself Before Shipping. It Found Five Omissions. | The build story — three rounds of refinement that missed what one recursive audit caught, and the architectural moves that made it work |
| I Asked for an Eker-Style Rewrite “Just for Fun.” It Became a Permanent Voice Profile. | How a throwaway request became the Grounded Challenger voice — decoupling mechanics from substance, building a selector, and the Dignity Test that keeps the voice from drifting |
About the session
Origin: June 1, 2026. One build session: the Critical Partner skill and the Grounded Challenger voice profile were built in the same sitting.
The two artifacts are related but distinct. The Critical Partner skill is about how AI evaluates your thinking. The Grounded Challenger voice profile is about how you sound when you write. The session produced both because the work on the skill surfaced an Eker rewrite that led somewhere unexpected.
Related skill
The Critical Partner skill referenced in these articles lives in wiki/mastermind/skills/ (session-extracted IP). The Grounded Challenger voice profile is in wiki/mastermind/skills/ under brand-writing resources.